George Bilbrey over at ReturnPath blogged about a new white paper published on the conversion and delivery rates of three symbolic spam campaigns. Since he always has great insight to the analytical nature of email, I had to read it.
Disclaimer: this is a very technical and empirical overview of the lower level protocols measured when sending email. You could probably just skip to the conclusion if you want to bypass the framework of the experiment and get to the point, although I would highly suggest reading the entire paper to spur ideas on how we in the email industry can measure the affect of spam on our email at ISPs. While none of us reading this blog presumably send blatant spam, we are directly impacted by the defenses ISPs put up and constantly change to thwart new spam attack vectors. This directly impacts email deliverability.
A few thoughts:
- The team who executed this test explained how their approach was ethical. By only acting as a proxy, not originating any new spam email (but rather capturing mail already in flight) and rewriting URLs to be benign, they side stepped a potentially dangerous situation. From my experience, granted it’s nothing compared to the researchers who have a lot more education than me, punishing recipients or introducing a punitive bias doesn’t work in the business world. It just pisses people off if and when they find themselves at the receiving end of an experiment they didn’t opt in to and can run afoul of the law in this case.
- No authentication was tested as an additional variable to the spam delivery rates. It would be interesting to see how the presence of DK/DKIM/SPF/SID would affect this.
- No open rates were measured. It would seem that if links could be rewritten in the message body, spacer images could be inserted to indicate an upper bound of emails being rendered on client email programs.
- The CBL (Composite Blocking List) was heavily monitored during the time frame to see when blacklisting occurred and at what rate. It would be interesting to measure other RBLs as well to show the efficacy of one v. another.
- Finally, the presumption that the TLD (or ending extension of the recipient’s email address such as .au or .com) shows where the recipient is actually located is a bit limiting. I’m not sure how you could avoid this, other than some sort of geo IP mapping with the calling IPs for links/images, but there are definitely recipients who get an email address at one regional ISP but take that with them when they move to another region. For instance, in my experience with EU addresses, there is a tendency for recipients to get a .com (usually US) address regardless of where they are located. Although a small segment, this would directly correlate to the causation of the spam education affecting lower spam conversion.
Net of the paper is that while spam might be a revenue generator, the amount of profit derived from the spam campaigns tested (pharmacy, post card and April Fools) is driven down by the amount of expenses associated with keeping an email infrastructure in place. The cost of hosting bad domains over a period of time and setting up/maintaining alternate routes of detection is considerable.
This is good news for legitimate senders in that spam doesn’t directly cause one to become rich and thus decreases the appeal to entering into the business in the first place. Or at least I hope.
Last 5 posts by Chris Wheeler
- What the !#$@ do I do with the headers? - April 19th, 2010
- Where the !#$@ are the headers? - March 29th, 2010
- Deliverability Forum: It's a Wrap! - January 11th, 2010
- Cultural Bias: A Discussion Around ISP and ESP Relations - December 17th, 2009
- Future of Deliverability: A Series - November 16th, 2009
